One of the central tenets of informed consent, as identified by Faden and Beauchamp, is the idea of autonomous actions on the part of the person making a consent decision. An autonomous action is that which is performed intentionally, with understanding, and without controlling influences. According to them, and as will be explained in more detail, expressions such as ``respect for autonomy'' and ``exercise of autonomy'' are often used in medicine in the context of informed consent without examining whether the expectations of the autonomy involved are aspirational or realistic (that is, whether the criteria for an action to be autonomous are actually practically applicable), or what the defining characteristics of an autonomous action or person are. Much debate has been had over the latter issue, with responses such as Benn's (1976) aspirational ideal theory that makes a distinction between what he calls ``normal choosers'' from ideal choosers. An ideal chooser is not affected by social pressure or convention, but develops ``a set of beliefs, values, and principles, by which his actions are governed''. This account is considered aspirational by Faden and Beauchamp because if held to such high standards, most autonomous-seeming decisions would not be classified as autonomous.
There are further challenges in applying theories of autonomy of persons to informed consent, the largest of these being the problem whereby a fully autonomous person may not necessarily act autonomously all the time. They may have the capacity for acting autonomously, and therefore be considered autonomous, but this is somewhat different from acting autonomously. There could be various reasons for an autonomous person to act non-autonomously, such as in temporary illness, or if the person is overwhelmed, ignorant, or doesn't care, to name a few possible situations. Such cases are particularly important for informed consent because any of these situations could result in an uninformed decision and thus invalidate the informed consent decision. These situations are extremely common in information technology1.3, and are particularly difficult to test for firstly because of the lack of general acceptance of the characteristics of autonomous persons, and secondly because of the limitations of technology.1.4
Beauchamp, in another book with Childress (1994) further qualify autonomous action by stating that ``an autonomous person who signs a consent form without reading or understanding the form is qualified to act autonomously by giving an informed consent, but has failed to do so'' (p. 121). This is a common problem for informed consent in information technology, as shown by End User License Agreements (section 2.1), with computer users rarely reading let alone understanding the contracts before agreeing to them.
For these reasons, Faden and Beauchamp (and Beauchamp and Childress) restrict their discussion of autonomy to that of actions rather than people. They state that they are interested in ``what an informed consent is - rather than problems of from whom consent should be or may be obtained'' (p. 237) and ultimately what is required to enable a person to make a sufficiently autonomous decision. This is an interesting distinction to make for their purposes, because although they discuss competence at a later point in the text, which brings up many similarly difficult arguments, it allows them to put aside the controversial discussion about what qualifies as an autonomous person, because it doesn't affect the main thrust of their argument. It should be noted that in moral philosophy and legal theory, an apparently ``freely'' chosen action may not be so, precisely because the agent fails a test for autonomy. It seems, though, that Faden and Beauchamp are aware of this, because they only fully discuss those who would be considered autonomous who then act non-autonomously (such as signing a form without understanding it), as opposed to non-autonomous agents that might act autonomously (or not). Those who would be considered non-autonomous in the moral philosophical or legal sense are merely tested for their capacity for an autonomous consent action by Faden and Beauchamp. However, even if a non-autonomous agent passes the capacity test, it may still fail to give informed consent because the agent itself fails a test for autonomy. Faden and Beauchamp are still right to talk about the autonomy of the action, but it must be understood that within their theory there is a connection between autonomous actions and the autonomy of the person. This problem ultimately increases the complexity of determining the autonomy of an action, since some sort of personal autonomy ultimately needs to be considered for informed consent, if autonomy is the foundation and focus of the theory.
According to Faden and Beauchamp, an autonomous informed consent action manifests itself as an ``autonomous authorisation'', which consists of three parts. As stated earlier, an autonomous action must be made intentionally, with understanding, and without controlling influences. Intention and understanding are necessary but not sufficient conditions for an autonomous action (not sufficient in that they could easily fail to account for temporary incapacity or mental illness, or would fail in a situation with controlling influences). The criterion of ``without controlling influences'' generally refers to external controlling influences inflicted upon the person as opposed to something inherent in the person, such as mental illness. The criteria of understanding and control can be gauged by degrees. An example that illustrates this concept is a ``gun to the head'' type example, where a victim at a robber's gunpoint may well understand exactly what is happening and the implications of making a decision, and may make it intentionally, but such a decision would not be autonomous because it is controlled by the robber. However, when grey areas emerge, such as when there is not complete understanding or control, it becomes more important to assess the degrees of these criteria and decide as to whether they pass a particular threshold test for autonomy. For example, in situations where the details needed for understanding are extremely complex, such as in a complicated medical procedure that needs to be performed quickly, the consenter is likely to be under a considerable amount of stress and (depending on the situation) possibly pain. The patient therefore may not necessarily understand the implications of a simplified or rushed explanation, or may just wish to defer to the doctor who seems to know more about what will make the consenter feel better. In this way the doctor may be seen to control the patient, especially if they use language laden with loaded terms, such as ``you will die if we don't operate'', but it is not a complete control as with the ``gun to the head'' example. As it is practically impossible to gain full autonomy, Faden and Beauchamp suggest that this threshold should be the ``general goal of substantially autonomous action.'' Beauchamp and Childress further clarify this position by stating that ``for an action to be autonomous we should only require a substantial degree of understanding and freedom from constraint, not a full understanding or a complete absence of influence'' [Beauchamp & Childress, 1994]. What is considered substantial is context-specific and difficult to define, however it can be reasonably established by considering the degrees of understanding and controlling influences. For example, it might be considered sufficient for a surgery patient to know they are having a gall stone removed via keyhole surgery, but not need to know the painstaking details by which the removal will take place, for a substantially autonomous action to take place. It might not, however, be sufficient for the patient to simply know they are having their cancer treated without understanding the side effects or implications of the chemotherapy they will undertake. In terms of the End User License Agreement examples discussed extensively in chapters 2 and 3, this allows for vague minimal requirements that could be easily exploited through ambiguous wording. Without some sort of regulatory body to enforce such minimum requirements, it is left to the manufacturer to decide how to interpret the idea of what constitutes ``substantial''. There is also no test for capacity within this framework, and it is troubling to concentrate on testing the conditions required for this autonomous authorisation without testing the capacity of the agent making the decision: testing these requirements only makes sense, after all, if there are agents involved, and these agents are autonomous agents.
Faden and Beauchamp are concerned with a general problem in the literature on informed consent where it is implied that only when a patient fully understands and is truly independent does their decision qualify as informed consent. They claim that these goals of full understanding and true independence are idealistic and virtually impossible to realistically achieve when creating a practically achievable theory. These goals for informed consent have merely served to attempt to achieve a substantial difference in autonomy of consent decisions, that is, they have generally improved informed consent situations and educated practitioners and the public as to the importance of informed consent, but have not been used as a threshold for demarcating informed from uninformed consent decisions. The idea of a substantially autonomous action threshold has instead been used, with the judgement of its being met dependent on context, and it is this that Faden and Beauchamp identify and address in their theory of informed consent as an autonomous action.
I will now discuss the three conditions of an autonomous action, according to Faden and Beauchamp: intention, understanding, and noncontrol, and their application to the idea of informed consent as an autonomous authorisation.